Leveraging replication to improve disaster recovery

Replication is becoming more affordable, thanks in part to host- and network-based replication as well as ATA and Serial ATA disk subsystems and new software.

By Karen Dutch

Although traditional disaster-recovery challenges (i.e., natural disasters, fires, massive power outages) loom as large as ever, organizations that are preparing disaster-recovery strategies today must address a host of new challenges and concerns. Fortunately, new software technologies are now available to help augment current infrastructures to keep business applications up and running.

Recent regulatory mandates such as HIPAA and Sarbanes-Oxley require organizations to ensure their data remains secure under any circumstances. In other cases, intensive global competitive pressures are forcing organizations to protect their data in ways that will enable them not only to recover from any disaster but also to do it fast—within hours or minutes.

An outage of just a few hours could result in significant lost business and squandered opportunities. According to research analysts, hourly losses from system outages can range from $300,000 to more than $3 million.

Finally, organizations are under severe pressure to reduce costs. The labor alone entailed in traditional disaster-recovery strategies, particularly in tape handling, is too high and the results are too slow and unreliable. Faced with these kinds of pressures, organizations cannot afford to rely on the disaster-recovery strategies of the past.

Companies worldwide are spending more than $1 billion a year collectively on data replication, according to the Gartner IT consulting firm.

This ranks replication as the second-highest line item for IT storage software spending, closely behind conventional backup and recovery ($1.5 billion).

Traditional disaster recovery—defined primarily as either tape backup or selective mirroring—is no longer adequate for some companies. Full tape backups are slow, and mirroring can drive up costs far exceeding the value of the data.

Organizations typically make full tape backups of their critical data weekly and incremental nightly backups of changed data.

Tapes are then shipped off-site to a secure facility. In the event of a disaster, people scramble to fetch the tapes and bring them to a disaster-recovery site where they mount the tapes and recover applications and data.

This process can take hours or even days. When remote systems finally become operational, they are running with data that may be a day or more old.

New methods of replication

Fortunately, organizations today have a better option. Data replication has emerged as a practical DR strategy for organizations of all sizes. New advances, such as host- and network-based replication, now bring data replication within the reach of any organization.

In the past, data replication was not practical for many organizations, especially for disaster recovery. For one thing, it was limited to short distances, often within a building or across a metropolitan campus. Today, however, organizations need to replicate data over distances of 100 miles or more to be confident that the disaster-recovery site will be outside the impacted area.

Cost was also a big problem. Data replication required that organizations invest in two costly storage infrastructures. Also, the organization often found itself mirroring data that didn't need to be mirrored at all, further driving up costs.

Recent advances have combined to make data replication for disaster recovery practical for almost any organization. New advances include the following:

  • IP networks provide a ubiquitous, low-cost mechanism to transmit data almost anywhere in the world.
  • Asynchronous replication offers a way to transfer data over long distances without impacting application performance.
  • Low-cost disk arrays allow organizations to assemble large quantities of inexpensive disk capacity.
  • New replication alternatives, such as host- and network-based replication, give organizations more flexibility.
  • Intelligent automation provides improved, policy-driven storage management tools that reduce the likelihood of error and improve consistency and reliability.

As a result, organizations can use data replication to expand the amount of critical data that is replicated. IT administrators can selectively mirror data across long distances, enabling disaster-recovery sites to be located hundreds or thousands of miles from the primary location.

Combining asynchronous replication with frequent point-in-time snapshot copies, organizations can achieve nearly real-time data currency without impacting applications. In the event of a disaster, only a few transactions might be lost or have to be reconstructed.

Data replication also removes the constraint of the overnight backup window, which was proving increasingly problematic in 24x7 IT operations. Additionally, the latest replication and storage management tools eliminate the errors and labor-intensive effort that plague tape backup, while supporting heterogeneous storage environments.

Given the new replication alternatives, organizations can select the most appropriate and cost-effective replication based on their recovery-time objectives.

Disaster recovery runs much smoother in a data-replication scenario. When disaster strikes, the primary systems fail, automatically passing processing chores to the remote site, where applications and databases are brought up within minutes. The mad scramble to find and mount tapes, and inconsistent restoration of outdated data, is eliminated.

However, tape is not going away. It will continue to be used for archival purposes; but even here replication will pay dividends. Organizations can make archival copies on tape from a replicated copy, which eliminates any impact on production systems.

Ultimately, data replication changes disaster recovery from a risky or costly proposition to a reliable business process. Advances in replication technologies, falling disk prices, ubiquitous IP networks, flexible replication options, and advanced replication and storage management software enable replication to meet DR challenges.

Karen Dutch is VP of product management at Softek (www.softek.com) in Sunnyvale, CA.

This article was originally published on December 01, 2003